Compressed Sensing Based SpO2 and PR Estimation for Improving Resource Utilization Efficiency of Wearables in IoMT Applications

Continuous measurement of oxygen saturation (SpO2) and pulse rate (PR) is most essential for diagnosis of cardiovascular and chronic pulmonary diseases. Affordable miniaturized devices with limited battery capacity highly demand energy-efficient continuous vital sign measurement for long-term health and wellness monitoring. In this paper, we present compressed sensing (CS) based SpO2 and PR estimation method by directly processing compressive PPG measurements obtained by using deterministic binary block diagonal (DBBD) sensing matrix. The performance of CS-PPG is studied by using four SpO2 estimation equations with the duration of 2 and 3 seconds PPG signals. Results show that linear equation provides accurate SpO2 measurement with less estimation error in terms of mean absolute error (MAE) of 0.69% and Bland Altman ratio (BAR) of 0.94% as compared to that of the quadratic equation based SpO2 measurement. Evaluation results of PR estimation approach show that the average magnitude difference function (AMDF) based approach had the best estimation error range (EER) metric having more number of PPG segments with zero error and error less than 5 bpm and MAE of 2.94 bpm for red-wavelength PPG as compared to the fast Fourier transform (FFT) based PR estimation approach having MAE of 3.72 bpm for 10 second PPG segment. From this study, the proposed CS-PPG based SpO2 and PR estimation method can reduce computational load and overall energy consumption by a factor of 5 as compared to the uncompressed PPG based measurement method.